nablaNABLA: Neighborhood Adaptive Block-Level Attention
Paper • 2507.13546 • Published • 126
Andrey Goncharov
aigoncharov
AI & ML interests
DL interpretability and performance
Organizations
upvoted a paper 9 months ago
upvoted 5
papers about 1 year ago
Feature-Level Insights into Artificial Text Detection with Sparse Autoencoders
Paper • 2503.03601 • Published • 233
When an LLM is apprehensive about its answers -- and when its uncertainty is justified
Paper • 2503.01688 • Published • 21
Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach
Paper • 2502.05171 • Published • 154
Your Transformer is Secretly Linear
Paper • 2405.12250 • Published • 157
SynthDetoxM: Modern LLMs are Few-Shot Parallel Detoxification Data Annotators
Paper • 2502.06394 • Published • 89